Electrical spin injection and detection in a semiconductor. Is it feasible?

Abstract

The electrical injection of spin polarized electrons in a semiconductor can be achieved in principle by driving a current from a ferromagnetic metal, where current is known to be significantly spin polarized, into the semiconductor via ohmic conduction. For detection a second ferromagnet can be used as drain. We studied submicron lateral spin valve junctions, based on high mobility InAs/AlSb two-dimensional electron gas (2DEG), with Ni, Co and Permalloy as ferromagnetic electrodes. In the standard geometry it is very difficult to separate true spin injection from other effects, including local Hall effect, anomalous magnetoresistance (AMR) contribution from the ferromagnetic electrodes and weak localization/anti-localization corrections, which can closely mimic the signal expected from spin valve effect. The reduction in size, and the use of a multiterminal non-local geometry allowed us to reduce the unwanted effects to a minimum. Despite all our efforts, we have not been able to observe spin injection. However, we find that this 'negative' result in these systems is actually consistent with theoretical predictions for spin transport in diffusive systems.